LINEs Contribute to the Origins of Middle Bodies of SINEs besides 3' Tails.

Short interspersed elements (SINEs), which are nonautonomous transposable elements, require the transposition machinery of long interspersed elements (LINEs) to mobilize. SINEs are composed of two or more independently originating parts. The 5' region is called the "head" and is derived mainly from small RNAs, and the 3' region ("tail") originates from the 3' region of LINEs and is responsible for being recognized by counterpart LINE proteins. The origin of the middle "body" of SINEs is enigmatic, although significant sequence similarities among SINEs from very diverse species have been observed. Here, a systematic analysis of the similarities among SINEs and LINEs deposited on Repbase, a comprehensive database of eukaryotic repeat sequences was performed. Three primary findings are described: 1) The 5' regions of only two clades of LINEs, RTE and Vingi, were revealed to have contributed to the middle parts of SINEs; 2) The linkage of the 5' and 3' parts of LINEs can be lost due to occasional tail exchange of SINEs; and 3) The previously proposed Ceph-domain was revealed to be a fusion of a CORE-domain and a 5' part of RTE clade of LINE. Based on these findings, a hypothesis that the 5' parts of bipartite nonautonomous LINEs, which possess only the 5' and 3' regions of the original LINEs, can contribute to the undefined middle part of SINEs is proposed.

Introduction

Short interspersed elements (SINEs) are composite mobile elements that can mobilize dependent on the help of counterpart long interspersed elements (LINEs), also called non-long terminal repeat (non-LTR) retrotransposons (Kajikawa and Okada 2002; Dewannieux et al. 2003). SINEs are composed of several independent parts: a head, body, and tail.

The heads of SINEs typically originate from noncoding RNAs such as 7SL RNA, tRNA, and 5S rRNA, which are the key for one classification scheme of SINEs (Kapitonov and Jurka 2003). SINEs with 7SL RNA-derived heads are called SINE1 and are only found in Euarchontoglires (primates, tree shrews, and rodents) (Kriegs et al. 2007). SINEs with tRNA-derived heads are the most widely distributed among eukaryotes and are called SINE2 (Bao et al. 2015). SINEs with 5S rRNA-derived heads are called SINE3 (Kapitonov and Jurka 2003). These SINEs are transcribed by RNA polymerase III depending on the activity of internal promoters inside of these SINE heads. Recently, a new group of SINEs with U1 or U2 snRNA-derived heads was proposed and designated SINEU (Kojima 2015). SINEs with 28S rRNA-derived sequences (SINE28) and with GC-rich sequences of unknown origins have also been proposed (Longo et al. 2015; Suh et al. 2016). High copy numbers of these newly proposed SINEs with nearly identical structures suggest that they are retrotransposition units, and not chimeric copies derived from two or more RNA templates, although the transcription mechanism for these SINEs are not yet demonstrated.

The 3′ termini of SINEs are called tails and are responsible for the mobilization of SINEs. Tails often exhibit sequence similarity to 3′ regions of LINEs, and the secondary structure of the tail is recognized by the proteins encoded by LINEs (Ohshima et al. 1996; Kajikawa and Okada 2002). However, many SINEs, represented by Alu elements from primates, do not have 3′ tail shared by their counterpart LINEs. In the case of Alu, the counterpart LINE, LINE-1 (L1), can mobilize any RNA with 3′ polyA tail, even mRNAs (Dewannieux et al. 2003). Such “relaxed” recognition of RNA by L1 proteins is likely the key of the success of many mammalian SINEs without specific 3′ tail sequences (Okada et al. 1997; Ohshima 2012).

The simplest SINEs, such as B1 and ID from rodents, have only a head and a 3′ tail. Some SINEs contain additional sequences unrelated to either LINEs or small RNA genes between the head and the tail. Among the three parts constituting SINEs, the middle part (i.e., the body) is the most enigmatic. This region rarely exhibits any sequence similarity to anything but SINEs. It is of interest because SINEs from very divergent animals sometimes exhibit significant similarity in the body region. Based on this similarity, several groups of SINEs, such as CORE-SINE, V-SINE, and Ceph-SINE, have been proposed (Gilbert and Labuda 1999; Ogiwara et al. 2002; Nishihara et al. 2006, 2016; Akasaki et al. 2010; Piskurek and Jackson 2011; Luchetti and Mantovani 2013). Even these body parts can be composite; sometimes SINEs share just the 5′ half of the body (Piskurek and Jackson 2011; Luchetti and Mantovani 2013; Nishihara et al. 2016).

Although the origins of widely conserved SINE bodies are completely unknown, the middle regions of some narrowly distributed SINEs have been characterized. One major group exhibits a bipartite structure of sequences that originate from LINEs. The bipartite structures often originate from 5′- and 3′-UTR of RTE-type LINEs. Examples include Bov-tA, Mar-1, AfroSINE, Ped-1, Ped-2, BuceSINE, GymnSINE, ManaSINE, and MeloSINE (Okada and Hamada 1997; Gilbert and Labuda 2000; Nikaido et al. 2003; Gogolevsky et al. 2008; Suh et al. 2016). The 3′ part originates from the extreme 3′ end including a 3′ polyA or microsatellite tail. The 5′ part is either the extreme 5′ end or an internal sequence inside of the 5′-UTR. We previously reported another type of LINE that can contribute to the bipartite structures of SINEs; the middle and 3′ terminal regions of SINE2-1_ACar and SINE2-1B_ACar exhibit similarities with the 5′ and 3′ of Vingi-2_ACar (Kojima et al. 2011).

Several nonautonomous LINEs possessing only the 5′ parts and 3′ parts of autonomous LINEs have been reported (Bringaud et al. 2003, 2009; Kojima et al. 2011). Their representatives are RIME derived from Ingi and NARTc derived from L1Tc (Bringaud et al. 2003, 2009). Vingi-1_EE have many nonautonomous derivatives generated due to internal deletion (Kojima et al. 2011). A proposed ancestral retrotransposition unit Bov-A, which is the shared part between Bov-A2 and Bov-tA, is an internally deleted derivative of the Bov-B LINE (Okada and Hamada 1997). Bov-A2 is a dimer of two Bov-A units, and Bov-tA is a combination of a tRNA-derived head and Bov-A. These observations—that is, the presence of nonautonomous LINEs with a bipartite structure and SINEs with a bipartite structure plus a 5′ RNA-derived head—raised the possibility that the middle parts of SINEs can originate from a part of LINEs. Here, this hypothesis is expanded to indicate the body of SINEs can be originated by bipartite LINEs even if SINEs do not have bipartite structures.

In this study, systematic analysis of the similarity between SINEs and LINEs and in-between is performed. Several new examples of bipartite structure of RTE-type LINEs in SINEs were found. A fragment of an RTE-derived sequence contributes to the latter half of the proposed Ceph-domain of SINEs, supporting the hypothesis that the conserved bodies of SINEs can be generated by a part of LINEs.

Materials and Methods

Repeat Detection and Classification

Multicopy sequences in published eukaryotic genomes were screened using approaches similar to those described previously in the literature (Bao and Eddy 2002). Screening for low-copy-number repeat sequences was also performed by Censor search (Kohany et al. 2006) with the protein sequences of well-characterized repeat sequences deposited in Repbase (Bao et al. 2015) (http://www.girinst.org/repbase). Classification is based on the similarity to known repeat sequences deposited in Repbase with Censor (Kohany et al. 2006). RTclass1 (Kapitonov et al. 2009) was used to further classify LINEs. All of the repeat sequences detected here have been deposited in Repbase. The similarity between LINEs and SINEs were analyzed with Censor and was confirmed via manual inspection. Sequence alignment was performed using MAFFT (Katoh et al. 2005) and MUSCLE (Edgar 2004) and was visualized using Jalview (Waterhouse et al. 2009) and UGENE (Okonechnikov et al. 2012).

Results

The Contribution of Bipartite LINEs to SINEs

The similarity between LINEs and SINEs and between different SINEs was analyzed using Censor with redundant option (Kohany et al. 2006). Censor used BLAST to compare the SINE sequences extracted from Repbase to the LINE sequences or the SINE sequences extracted also from Repbase. All LINE–SINE pairs and SINE–SINE pairs showing sequence similarity detected by Censor were extracted and inspected manually to remove accidental hits. First, the hits on the complementary strand were all removed. Several accidental hits were observed when a LINE had a low-complexity sequence (e.g., the sequence 6897–6977 of L1-10_PI). The presence of a tRNA-like sequence in RTE-1_DAn and its relatives results in hits between these LINEs and many SINE2 elements. After removing these hits, the remaining LINE–SINE pairs were analyzed to determine whether the LINE-derived sequences were present in the counterpart SINE besides the 3′ terminus. Because the similarity between LINEs and SINEs at their 3′ termini is common if the SINE is dependent on the transposition machinery of the LINE, this step is essential. Finally, SINEs that have been already reported to possess bipartite LINE structure (Bov-tA, Mar-1, AfroSINE, Ped-1, Ped-2, PlatSINE1, Plat_RTE1_SINE, BuceSINE, GymnSINE, ManaSINE, and MeloSINE) were removed (Okada and Hamada 1997; Gilbert and Labuda 2000; Nikaido et al. 2003; Gogolevsky et al. 2008; Bao et al. 2015; Suh et al. 2016). Goat NLA repeat is likely a member of Bov-tA. The structure, sequence, and distribution of SINE2-1_Laf from the African elephant Loxodonta africana and SINE2-1_Pca from the rock hyrax Procavia capensis suggest that they are members of AfroSINEs. RTESINE1 and RTESINE2 are both bipartite RTE-type nonautonomous LINEs. The final candidates for new bipartite LINE-derived regions seen in SINEs are shown in figure 1 and listed in table 1. The sequences of these SINEs along with information of their composite structure appear in supplementary figure S1, Supplementary Material online, and the alignments between LINEs and SINEs appear in supplementary figure S2, Supplementary Material online.

. 1.

—Schematic representation of SINE structures. The origins of head (tRNA or 5 S rRNA), body (CORE), and LINE-derived parts (RTE 5′-UTR, RTE 3′-UTR, Vingi 5′-UTR) are indicated. Regions whose origins are unknown are indicated by “?.”

Table 1

SINEs Whose Two Parts of Sequences Show Similarity to LINEs

SINE	Region	LINE	Region	Identity
CoeSINE4 (201)	84–134	RTE-4_PPo (3095)	53–107	81%
	141–194	RTE-4_PPo	3040–3093	76%
CoeSINE5 (225)	117–168	RTE-2_MMa (3176)	115–166	83%
	173–216	RTE-2_LVa (5082)	5034–5077	76%
HaSE1 (385)	255–311	RTE-2_DPl (3242)	188–243	77%
	341–385	RTE-N1_ATr (195)	144–186	91%
HaSE2_DP (299)	194–229	RTE-N2_Lch (286)	151–186	83%
	242–299	RTE-3_PXu (565)	504–558	84%
HaSE3 (349)	235–288	RTE-2_DPl (3242)	188–239	80%
SINE2_Gav (271)	78–223	RTE-11_AMi (3664)	295–439	87%
	223–267	RTE-11_AMi	3616–3659	89%
SINE2-1_EC (407)	85–302	RTE-1_OAf (3275)	1–221	88%
	317–406	RTE-1_OAf	3182–3269	83%
SINE2-1_PPo	199–277	RTE-N1_Lch (262)	152–233	78%
SINE2-1_PXu	197–296	RTE-N1_Lch (262)	152–250	73%
SINE2-2_ACar (239)	10–220	MAR1a (250)	6–246	69%
SINE-1_Pmo (241)	16–107	5S-Sauriaa (348)	29–119	87%
	134–239	BOVA2a (269)	1–122	74%
WALLSI1 (387)	170–335	RTE-1_PSi (3769)	27–196	69%
	338–378	RTE-1_PSi	3726–3767	86%
WALLSI1A (610)	415–564	RTE-3_AMi (3899)	147–292	74%
	570–609	RTE-3_AMi	3854–3893	83%
WALLSI2 (321)	134–266	RTE-3_AMi (3899)	159–289	74%
	275–317	RTE-3_AMi	3854–3895	88%

Note.—If the same region of SINE hits several different LINEs, only the LINE with the highest CENSOR score is shown. The length of LINE/SINE is shown in parenthesis.

SINEs originated by the internal deletion of LINEs.

CoeSINE4 and CoeSINE5

Two coelacanth SINE families, CoeSINE4 and CoeSINE5, have similar 3′ sequences (table 1). These sequences correspond to the 5′- and 3′-UTR of RTE-type LINEs. CoeSINE4 has a tRNA-derived head, and CoeSINE5 has a 5S rRNA-derived head.

HaSE1, HaSE2_DP, SINE2-1_PXu, and SINE2-1_PPo

HaSE1 and HaSE2 were reported from a lepidopteran insect Helicoverpa armigera by Wang et al. (Wang et al. 2012). HaSE2_DP is a HaSE2-related SINE from another lepidopteran insect, the monarch butterfly Danaus plexippus. The 5′ ∼130-bp sequences of HaSE1 and HaSE2_DP are 78% identical, and this region corresponds to the 5′ tRNA-derived head and “conserved central domain” reported by Wang et al. (Wang et al. 2012). SINE2-4_NV from sea anemone exhibits similarity to both 5′ regions of HaSE1 and HaSE2_DP. Furthermore, HaSE2_DP exhibits sequence similarity to two butterfly SINEs (SINE2-1_PXu and SINE2-1_PPo) with the exception of the 5′ half of the tRNA-derived region. The alignment of these SINEs with SINE2-5_NV, which is also similar to SINE2-4_NV, reveals the strong similarity among HaSE2_DP, SINE2-1_PXu, and SINE2-1_PPo starting around nucleotide 130 of HaSE2_DP (data not shown). In contrast, the 3′ region of HaSE1 is similar to SINE2-5_NV. However, the “conserved central domain” does not exhibit strong conservation among these six SINE families.

The sequence 255–311 of HaSE1 exhibits similarity with the 5′-UTR of the autonomous RTE-type LINE from the monarch, RTE-2_DPl (table 1). The 3′ end of HaSE1 was reported to be similar to the 3′ end of RTE-3_BM from the domestic silkworm Bombyx mori (Wang et al. 2012). A Censor search with Repbase yields a more similar sequence in RTE-N1_ATr from a plant Amborella trichopoda, but the sequence similarity is restricted to the ∼40-bp 3′ end (supplementary fig. S1, Supplementary Material online).

The 3′ regions of HaSE2_DP exhibit similarity to RTE-N2_Lch from coelacanths (table 1). RTE-N2_Lch is an internally deleted derivative of RTE-4_Lch. RTE-N2_Lch corresponds to the 5′ 214 bp and the 3′ 70 bp of RTE-4_Lch. Therefore, the similarity of HaSE2_DP to the region 151–186 of RTE-N2_Lch indicates that HaSE2_DP contains a sequence originating from the 5′-UTR of RTE. It is noteworthy that the 3′ region of HaSE2 has been reported to be from a Mariner DNA transposon (Wang et al. 2012). However, the presence of a sequence similar to RTE indicates that HaSE2 is also a canonical SINE whose 3′ region originates from a LINE. SINE2-1_PXu and SINE2-1_PPo also contain the sequence of bipartite RTE (table 1).

HaSE3 and HaSE1 share 3′ sequences but are different in their 5′ regions. Instead of a tRNA-derived head and conserved central domain of HaSE1, HaSE3 has a 5S rRNA-derived head. The 3′ end of HaSE3 is similar to that of HaSE1, and they share a common origin of the 3′ end of RTE (table 1).

WALLSI

Five WALLSI subfamilies (WALLSI1, WALLSI1A, WALLSI2, WALLSI3, and WALLSI4) have been reported from the tammar wallaby Macropus eugenii. WALLSI subfamilies other than WALLSI2 have also been found in the Tasmanian devil (Nilsson et al. 2012). The 3′ half of WALLSI1 is similar to that of MAR4_MD, a bipartite nonautonomous RTE from the opossum Monodelphis domestica. WALLSI1, WALLSI1A, WALLSI2, and WALLSI3 share very similar 3′ halves that exhibit strong similarity to the 5′- and 3′-UTRs of RTE (table 1). WALLSI3 has been revealed to be a bipartite nonautonomous RTE and is very similar to RTESINE2, an older bipartite nonautonomous RTE family, which is also found in the genome of the opossum M. domestica (Nilsson et al. 2010). The 5′ ∼130 bp of WALLSI2 is similar to the corresponding regions of the MIR and THER1 families. Therefore, WALLSI2 is composed of a tRNA-derived head (roughly 1–80), CORE (roughly 80–133), 5′ part of RTE (134–266), and 3′ end of RTE (275–317) (supplementary fig. S1, Supplementary Material online). The 5′ regions of WALLSI1 and WALLSI1A do not exhibit any similarities with other transposable elements (TEs), tRNAs or 5S rRNA. WALLSI4 does not exhibit sequence similarity with any other WALLSI SINEs in its 3′ region, but its 5′ region is similar to that of WALLSI2. This finding suggests that WALLSI2 was generated by the fusion of a 5′ region of WALLSI4 and a 3′ region of WALLSI1, WALLSI1A, or WALLSI3. RTESINE2 and WALLSI3 are very similar, and RTESINE2 is older than any WALLSI subfamilies, which indicates that WALLSI3 is the direct descendant of RTESINE2 in the wallaby lineage and that WALLSI1 and WALLSI1A are the derivatives of WALLSI3 with swapped 5′ regions.

SINE2_Gav

SINE2_Gav from crocodilians is 271 bp in length. It is composed of a tRNAGly-like head (roughly 1–70), a middle sequence (78–223) similar to the 5′-UTR of RTE-11_AMi and a tail (223–267) similar to the 3′-UTR of RTE-11_AMi (supplementary fig. S1, Supplementary Material online).

SINE2-2_ACar

SINE2-2_ACar is 239 bp in length. It is composed of the 5′ tRNA-derived head, a CORE-like middle sequence and two regions derived from the 5′- and the 3′-UTRs of an RTE-type LINE (supplementary fig. S1, Supplementary Material online). The downstream sequence from the CORE shows no similarity to known LINEs, but it is similar to RTE-derived regions of some SINEs including AFROSINE3 and MAR1. Many SINE2-2_ACar copies are roughly 85% identical to the consensus. The structure of SINE2-2_ACar is identical to that of MAR1, and it therefore may be a distant relative of MAR1.

SINE-1_Pmo

SINE-1_Pmo is a SINE3 family from the python Python molurus. Although the 3′ end (188–241) of SINE-1_Pmo has no closely related LINEs, it exhibits similarity with BovA2 (table 1). A comparison between SINE-1_Pmo and BovB (a family of RTE and the counterpart LINE of BovA2) revealed that SINE-1_Pmo includes the sequences originating from the 5′- and 3′-UTRs of RTE.

SINE2-1_EC and Its Descendants

SINE2-1_EC, which originated from the horse Equus caballus, is 407 bp in length and has a 3′ region (85–406) exhibiting >80% sequence identity to the 5′- and 3′-UTRs of RTE-1_OAf from the aardvark Orycteropus afer (table 1). Therefore, the structure of SINE2-1_EC resembles that of AfroSINEs even though horses are not Afrotherians. Upstream of this sequence (6–77) is a tRNAGlu-derived head based on the result of tRNAscan-SE (http://lowelab.ucsc.edu/cgi-bin/tRNAscan-SE2.cgi) and a Censor search in Repbase. Interestingly, the 5′ 115-bp sequence of SINE2-1_EC is almost identical to that of ERE4B, another SINE from the horse. As a consequence, the downstream sequence of the tRNA-derived head of ERE4B exhibited a pronounced similarity to the 5′ ends of RTE. The entire length of ERE4B is similar to ERE4, ERE1, ERE1B, and ERE1C—all from the horse—as well as CERE1 from the white rhinoceros Ceratotherium simum. These SINEs may have a chimeric origin between a SINE2-1_EC-like sequence contributing to the 5′ half and another LINE or SINE contributing to the 3′ half. There are no clues in terms of the counterpart LINE for ERE1, ERE1B, ERE1C, ERE4, or ERE4B.

Solo LINE-Derived Sequences in the Middle of SINEs

Ancient bipartite LINE-derived sequences may have been exchanged by newly acquired 3′ tails derived from another LINE. This situation can lead to a structure in which only the middle part of the SINE exhibits a similarity with the LINE. ERE4B is an example of such a chimeric SINE. Manual inspection of the Censor results noted above revealed several candidates for this type of chimeric SINE (fig. 1 and table 2). The alignments between SINEs and LINEs appear in supplementary figure S3, Supplementary Material online. Among them, ERE4B is described earlier. The 3′ end of MARE3 corresponds to the middle of 5′-UTR of RTE-14_Lch, suggesting the current consensus sequence of MARE3 is 3′-truncated.

Table 2

Internal Fragments of LINE 5′-UTRs Seen in the Middle of SINEs

SINE	Region	LINE	Region	Identity
ERE4B (185)	82–116	RTE1-N1b_LA (470)	2–37	92%
EbuSINE2 (370)	270–321	RTE-3_MD (3228)	129–176	86%
MARE3 (180)	101–178	RTE-14_Lch (3944)	220–298	69%
MonoRep87A (523)	391–455	RTE-3_PM (3975)	294–359	76%
MonoRep87B (537)	403–467	RTE-14_Lch (3944)	286–348	78%
SINE2-1_AMi (161)	61–127	Vingi-2_Gav (3128)	2–74	84%
IdioSINE2 (423)	130–367	RTE-2_Croc (4296)	259–486	75%
OegopSINE (370)	130–220	RTE-3_BF (4202)	325–414	79%
	225–281	RTE-12_AMi (3904)	182–237	74%
SepiaSINE (278)	127–213	RTE-3_BF (4202)	325–414	76%
Sepioth-SINE1 (292)	134–239	RTE-3_BF (4202)	325–423	79%
Sepioth-SINE2A (294)	133–238	RTE-3_BF (4202)	325–423	77%

Note.—If the same region of SINE hits several different LINEs, only the LINE with the highest CENSOR score is shown. The length of LINE/SINE is shown in parenthesis.

EbuSINE2

EbuSINE2 has been reported to be a family of Deu-SINEs with a tRNA-derived head (Nishihara et al. 2006). The sequence downstream of the Deu-domain (278–321) exhibits similarity with the 5′-UTR (129–176) of RTE-3_MD (table 2 and supplementary fig. S1, Supplementary Material online). Although the 3′ terminus of EbuSINE2 exhibits no sequence similarity with any TEs in Repbase, this region may be derived from the 3′-UTR of an unknown RTE.

MonoRep87A and MonoRep87B

MonoRep87A and MonoRep87B are two SINE families from the platypus Ornithorhynchus anatinus. Their consensus sequences start with a tRNA-like sequence and end with (CAT)n microsatellites, indicating that they are full-length sequences of SINE2. Although there is no sequence similarity to known LINEs or SINEs in their 3′ termini, the upstream sequences exhibit similarity with the 5′-UTR of RTE (table 2 and supplementary fig. S1, Supplementary Material online). The middle regions of these two SINE2 families are similar, but no close relatives have been found.

SINE2-1_AMi

SINE2-1_AMi is a SINE2 family found from Alligator mississippiensis. Just downstream of the 5′ tRNA-derived head, there is a sequence similar to the 5′ end of Vingi-2_Gav from the gharial Gavialis gangeticus (table 2 and supplementary fig. S1, Supplementary Material online). This example is the only newly identified SINE containing a fragment of a LINE other than RTE.

The 3′ Half of the Ceph-Domain Originates from RTE 5′-UTR

The 3′ halves of the “Ceph-domain” of SepiaSINE, OegopSINE, Sepioth-SINE1, and Sepioth-SINE2A are similar to several LINEs belonging to the RTE clade (table 2 and supplementary fig. S1, Supplementary Material online). A repeated Censor search with these 3′ halves of Ceph-domains and RTE sequences in Repbase revealed a well-conserved domain between RTE, RTE-derived SINEs as well as Ingi-3_AC and R4-1_ADi (fig. 2 and supplementary fig. S4, Supplementary Material online). RTEs from diverse animals—including vertebrates, echinoderms, annelids, arthropods, and cnidarians— contain this conserved domain. The hits included recently characterized LINEs and SINEs from birds (Suh et al. 2016). BuceSINE, MeloSINE, ManaSINE1, and ManaSINE2 are assumed to have originated independently, but they all contain sequences showing similarity with the 3′ half of the Ceph-domain. Some RTEs, such as AviRTE, contain two regions corresponding to this conserved domain in their 5′-UTRs. It is noteworthy that this conserved domain is not located at the 5′ end but rather in the middle of the 5′-UTRs.

. 2.

—Sequence similarity of Ceph-domains with some RTE LINEs. Nucleotides identical to those in representative Ceph-domains (SepiaSINE and IdioSINE2) are shaded. The positions of consensus sequences are shown in parentheses with their origins.

An unexpected finding was that Ingi-3_AC and R4-1_ADi, very distant LINEs from RTE, contained a sequence similar to Ceph-SINE and RTE. The RTE-like sequence in Ingi-3_AC (243–310) is in the latter half of 5′-UTR (1–450). Because Ingi-3_AC is a LINE from the California sea hare Aplysia californica, one species of mollusks, it is possible that the recombination between the Ingi LINE and the Ceph-SINE contributed to this sequence similarity. R4-1_ADi is from coral Acropora digitifera. The sequence similar to RTE is located at 135–179, in the former half of 5′-UTR (1–652).

Another Ceph-SINE IdioSINE2 Has a Different 5′-UTR Fragment of RTE

The 3′ half of the Ceph-domain of IdioSINE2 is similar to vertebrate RTE families such as RTE-2_Croc from crocodilians and RTE-2_LCh from coelacanths (table 1 and fig. 2). This RTE-like sequence is not similar to RTE-like sequences from other Ceph-SINEs. However, upstream of this region, IdioSINE2 contains a short, 23-bp RTE-like sequence (CCTCCAGCTATGGGTTAAATAGT) similar to that of other Ceph-SINEs. It corresponds to the 5′ terminal sequence of the RTE-like sequence from other Ceph-SINEs. Considering the occasional replacement of LINE-like 3′ terminal sequences in SINE evolution, IdioSINE2 was likely generated via tail replacement by another RTE LINE with a short 23-bp fragment of original RTE-like sequence remaining.

Similarity between the CORE-Domain and the 5′ Half of the Ceph-Domain

It is now clear that the 3′ half of the Ceph-domain derives from the 5′-UTR of RTE. What about the 5′ half of the Ceph-domain? The originally reported Ceph-domain was ∼150 bp long. Excluding the RTE-derived region, the 5′ ∼50-bp sequence is here redetermined as Ceph-domain. A Censor search in Repbase revealed that this 5′ half exhibits weak similarity with the CORE-domain (fig. 3). The CORE-domain exhibits a high sequence diversity, and the conserved region among all reported CORE-domains is only ∼25-bp long. The Ceph-domain shares 15 bp with the conserved CORE-domain. The conserved sequence CCTTGGG in the Ceph-domain is also present in the CORE-domain. Two CORE-SINEs from mollusks, SINE2-1_CGi from the Pacific oyster Crassostrea gigas and CALSINE3 from the California sea hare Aplysia californica, share a longer identical sequence with the Ceph-domain CCTTGGGAAAG. It is reasonable to consider that the Ceph-domain is a cephalopod- (or mollusk)-specific derivative of the CORE-domain that has experienced the loss of the 3′ half of the CORE-domain due to tail replacement by RTE.

. 3.

—Alignment of CORE-domains and Ceph-domains. Nucleotides identical to MIR CORE-domain is colored in red, whereas nucleotides identical to Sepioth-SINE1 but not to MIR CORE-domain are in yellow. RTE-like sequences in Ceph-SINEs are in lower cases.

Discussion

Bipartite Nonautonomous LINEs and the Birth of New SINEs

In this study, several new SINE families that have the 5′ and the 3′ parts of LINEs were found. Not a few nonautonomous LINEs with solely the 5′ and the 3′ parts of autonomous LINE counterparts have been created by internal deletion (Bao et al. 2015). They can be subclassified into two types: ORF1-absent and ORF1-present. HeT-A, HAL1, and Ag-Sponge can be members of bipartite nonautonomous LINEs even though they encode one protein corresponding to ORF1p (Pardue et al. 1996; Smit 1999; Biedler and Tu 2003; Bao and Jurka 2010). HeT-A and related elements were derived from the Jockey clade of LINEs, HAL1 were from the L1 clade, and Ag-Sponge were from the CR1 clade. The proteins encoded by these nonautonomous elements likely function to multimerize with the proteins encoded by autonomous counterparts and to enhance transposition (Rashkova et al. 2002). The necessity of generating ORF1p excludes the possibility that these protein-coding nonautonomous LINEs function as a source of SINEs; SINEs cannot encode a protein.

The distributions of bipartite ORF-absent nonautonomous LINEs in the classification of LINEs are very biased. Only four clades of LINEs—RTE, Ingi, Vingi, and R2—have been reported to produce bipartite nonautonomous LINEs (Bringaud et al. 2003, 2009; Kojima et al. 2011; Eickbush and Eickbush 2012; Bao et al. 2015). Ingi and Vingi are closely related clades of LINEs (Kojima et al. 2011). Here, only two clades of LINEs, RTE and Vingi, were revealed to contribute to the middle parts of SINEs. It is obvious that some SINEs are descendants of bipartite nonautonomous LINEs as proposed previously for Bov-tA (Okada and Hamada 1997). A striking example is WALLSI2. WALLSI2 is the recombinant between WALLSI4 and either WALLSI1, WALLSI1A, or WALLSI3. WALLSI3 is a bipartite nonautonomous LINE, and WALLSI1 and WALLSI1A are likely descendants of WALLSI3 or RTESINE2, the latter of which has an identical structure as WALLSI3 but is older. WALLSI2, as well as SINE2-1_ACar, has a CORE-domain upstream of the 5′ part of RTE. It is very likely that SINE2-1_ACar is also a recombinant of a bipartite nonautonomous LINE and an unknown SINE having a tRNA-derived head and a CORE-domain.

The 5′ sequences of RTE observed in SINEs are not always the 5′ ends. In contrast, bipartite nonautonomous LINEs usually possess the 5′ end of their original LINEs. The presence of a self-cleaving ribozyme at the 5′ terminus of some LINEs may be a cause of this distinction (Ruminski et al. 2011). Several RTE families are predicted to possess a self-cleaving ribozyme (Ruminski et al. 2011). Considering the structure of SINEs, which possess an RNA-derived head upstream of their LINE-derived parts, the presence of a self-cleaving ribozyme causes 5′-truncation. Six clades of LINEs, R1, R2, R4, RTE, Ingi, and LOA, were revealed to possess a self-cleaving ribozyme at their 5′ ends (Eickbush and Eickbush 2010; Ruminski et al. 2011; Sánchez-Luque et al. 2011). Among them, R1, R2, and R4 are target-specific LINEs (Kojima and Fujiwara 2003, 2004) and likely depend on the transcription of target ribosomal RNA genes. They accordingly need to cleave their 5′ ends to generate full-length transcripts (Eickbush and Eickbush 2003, 2010). Three clades, RTE, Ingi, and R2, generate bipartite nonautonomous LINEs. It is not yet known whether this tendency is caused by sampling bias or by specific requirements of transcription.

The generation of bipartite nonautonomous LINEs may also be related to different requirement of transcription initiation. Bipartite LINEs are very likely transcribed by RNA polymerase II, as is true for their counterpart autonomous LINEs. SINEs, on the other hand, are transcribed by RNA polymerase III. It is known that the 5′ extreme regions of LINEs are responsible for transcription (Takahashi and Fujiwara 1999). The cis-regulatory sequences for transcription by RNA polymerase II may contradict efficient transcription by RNA polymerase III.

Origins of Conserved SINE Bodies

Currently, the V-domain, CORE-domain, Deu-domain, Nin-domain, Ceph-domain, Inv-domain, Pln-domain, Snail-domain, and Meta-domain have been proposed as conserved SINE bodies (Gilbert and Labuda 1999; Ogiwara et al. 2002; Nishihara et al. 2006, 2016; Akasaki et al. 2010; Piskurek and Jackson 2011; Luchetti and Mantovani 2013; Matetovici et al. 2016). However, Nin-domain and Inv-domain have been reported to be variants or parts of Deu-domain. The Snail-domain and the Nin-domain show similarity at their 5′ ends. In this article, the originally proposed Ceph-domain (Akasaki et al. 2010) is revealed to be composed of two regions of independent origins: the CORE-domain and the 5′-UTR of RTE. Although the sequence similarity between the CORE-domain and Ceph-domain is marginal (fig. 3), the sequence diversity among CORE-SINEs can rationalize the classification of the Ceph-domain as a member of the CORE-domain (Gilbert and Labuda 1999).

Recent analysis has revealed that some SINE “superfamilies” share 5′ regions of their bodies but not 3′ regions. Nishihara et al. (2016) reported that two different types of 3′ regions of the CORE-domain are present, and they designated them CORE (original) and CORE2. The Inv-domain is similar to the Nin-domain and is combined with the 3′ flanking Pln-domain in Polyneopteran insects (Luchetti and Mantovani 2013). The Nin-domain and Snail-domain exhibit sequence similarity only in their 5′ regions (Matetovici et al. 2016). The fusion of two bodies, such as the Meta-domain and the Deu-domain, is also observed (Nishihara et al. 2016). These facts suggest that these proposed domains are not minimal functional units. The replacement of parts of the body appears common.

Here, a hypothesis that nonautonomous LINEs that have only 5′ and 3′ regions of original LINEs can be a source of enigmatic middle body of SINEs is proposed (fig. 4). This can be considered as an extension of the hypothesis by Okada and Hamada (1997), in which some SINEs originated from the addition of 5′ heads onto an internally deleted derivative of autonomous LINEs. Very limited groups of LINEs can generate internally deleted derivatives for unknown reasons. Such nonautonomous bipartite LINEs can be transcribed by RNA polymerase II and transpose dependently on the original autonomous LINEs. A template switch can add a 5′ small RNA-derived sequence onto a bipartite LINE, resulting in the birth of a SINE that is transcribed by RNA polymerase III. Due to the occasional exchange of parts of SINEs, the 5′ and 3′ regions of LINEs cannot always be present in combination in SINEs, which is demonstrated by the structure of ERE4B. Once the 3′ LINE-derived sequence is exchanged, characterizing the origin of the middle bodies of SINEs is a challenge due to their short lengths and relatively low sequence conservation compared with the rapid sequence evolution of mobile elements. The LINE-originated sequence in ERE4B is only 35 nucleotides in length. It would be nearly impossible to characterize the origin of this kind of short fragmented sequence if the counterpart LINE went extinct. This situation is perhaps why no sequence similar to conserved body sequences of SINEs has been found.

. 4.

—Hypothetical origin of SINE body.

SINEs which contain similar RTE 5′ regions, such as avian BuceSINE, ManaSINE, MeloSINE, and Ceph-SINEs, have independently evolved. A high sequence similarity of RTE 5′ regions between SINEs from diverse animals has been observed. For example, the RTE 5′ sequence from CoeSINE4 from coelacanths is ∼87% identical to that of SINE2-1_PPo from butterflies. This high sequence similarity resembles conserved SINE bodies. Conserved SINE bodies are often observed in conserved noncoding elements (Nishihara et al. 2006; Xie et al. 2006). They have been exapted to have a certain biological function, such as enhancer, promoter, or insulator (Bejerano et al. 2006; Sasaki et al. 2008). The ability to bind to a transcriptional regulator can also be useful for SINEs and LINEs, and it can accordingly be speculated that the conservation of the 5′-UTR sequences among diverse RTE LINEs as well as SINEs is due to their functional importance in the lifecycle of these mobile elements. Such functional elements can be maintained in evolution and are poised to become integrated into host biological systems.

Supplementary Material

Supplementary data are available at Genome Biology and Evolution online.

Click here for additional data file.